Zynerji/Ektome-Qwen2.5-1.5Bi-PristinelyUncensored

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 15, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Ektome-Qwen2.5-1.5Bi-PristinelyUncensored is a 1.5 billion parameter Qwen2.5-based language model developed by Zynerji. This model is specifically designed to be uncensored by isolating and removing only refusal-specific components, aiming to retain general helpfulness without capability degradation. It focuses on providing a fully compliant model while attempting to preserve the original model's performance, making it suitable for applications requiring unfiltered responses.

Loading preview...

Overview

Zynerji's Ektome-Qwen2.5-1.5Bi-PristinelyUncensored is a 1.5 billion parameter model based on the Qwen2.5 architecture, distinguished by its "Ektomē" process. This proprietary method aims to create an uncensored model by precisely excising only refusal-specific components, rather than broad abliteration that can degrade general knowledge and reasoning. The goal is to maintain the pristine model's capability while achieving full compliance against refusal behaviors.

Key Characteristics

  • Uncensored by Design: Utilizes a specific "Ektomē" process to remove refusal-specific components, ensuring the model will not refuse prompts.
  • Capability Retention Focus: The method is designed to preserve the original model's capabilities, with a reported MMLU-val of 0.578, identical to the pristine Qwen2.5-1.5B-Instruct.
  • Certified Compliance: Achieves 1.000 compliance on harmful content, indicating its uncensored nature.
  • Inconclusive Capability Certificate: While arithmetic and reasoning axes passed non-inferiority tests, instruction and knowledge axes were inconclusive due to sample size, meaning no definitive retention claim is made for these specific areas.
  • Quantizations Available: Provided in various GGUF quantizations, including IQ* variants for improved quality at lower precision.

Use Cases

This model is particularly suited for applications where:

  • Unfiltered Responses are Required: Ideal for scenarios where a model must not refuse to answer, regardless of prompt content.
  • Research into Uncensored LLMs: Useful for studying the effects of targeted refusal removal on model behavior and capability.
  • Specific Content Generation: Can be employed in creative or niche applications where content filtering might be undesirable, with the understanding that the user is accountable for its output.

Finetunes
1 models